Scenario 8 (More Generations): 2 Experiments With Different Learning Rates
Experiment 8.1
Time: 2016-04-15 16:00:28
Commit: 0e5134c2af678ddd626cce6e44d98086a9089981
Parameters
learningRate : 0.3
randomJump : 1
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 300
numberOfAgents : 150
bufferSize : 10
Benefit per unit of time
Experiment 8.2
Time: 2016-04-15 17:12:13
Commit: ff447f5b291c44063eec67a197b19afdba504578
Parameters
learningRate : 0.5
randomJump : 1
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 300
numberOfAgents : 150
bufferSize : 10
Benefit per unit of time
</div>
</body> </html>
Scenario 7: 24 Experiments Exploring Parameter Space
- Experiment 7.1
- Experiment 7.2
- Experiment 7.3
- Experiment 7.4
- Experiment 7.5
- Experiment 7.6
- Experiment 7.7
- Experiment 7.8
- Experiment 7.9
- Experiment 7.10
- Experiment 7.11
- Experiment 7.12
- Experiment 7.13
- Experiment 7.14
- Experiment 7.15
- Experiment 7.16
- Experiment 7.17
- Experiment 7.18
- Experiment 7.19
- Experiment 7.20
- Experiment 7.21
- Experiment 7.22
- Experiment 7.23
- Experiment 7.24
Experiment 7.1
Time: 2016-04-14 16:50:29
Commit: 0883dc6cbc273500d66c67227c6143182413fa55
Parameters
learningRate : 0.3
randomJump : 1
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.2
Time: 2016-04-14 19:24:49
Commit: 159e2da6c706aa2588955bc033b7aea7dc1e9ecc
Parameters
learningRate : 0.1
randomJump : 99
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.3
Time: 2016-04-14 20:31:15
Commit: 17d8194b65477a2c95a69d5fffa6491fcc391d82
Parameters
learningRate : 0.1
randomJump : 50
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.4
Time: 2016-04-14 18:16:04
Commit: 186e78a5ad721baac67496a49fe1e6c146f197bc
Parameters
learningRate : 0.1
randomJump : 1
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.5
Time: 2016-04-14 14:53:37
Commit: 26c5cbd7b38ec4baaf89a13cd2bdc217d6a0b836
Parameters
learningRate : 0.1
randomJump : 1
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.6
Time: 2016-04-14 15:28:44
Commit: 27ce18d5d7dac9cf5cf2d4be2c6da3376c598f16
Parameters
learningRate : 0.1
randomJump : 50
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.7
Time: 2016-04-14 19:41:51
Commit: 3226c69f03155d7baaf16718d63db25874ad70d6
Parameters
learningRate : 0.3
randomJump : 99
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.8
Time: 2016-04-14 15:58:20
Commit: 37e6f3861beadb8ea49161a5e16b9335cdab29ed
Parameters
learningRate : 0.1
randomJump : 99
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.9
Time: 2016-04-14 18:50:20
Commit: 3b875d850fb0e8dd7985a098dd7b13553139a070
Parameters
learningRate : 0.1
randomJump : 50
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.10
Time: 2016-04-14 16:14:45
Commit: 5f60873edb355644532dacc21f93dd2fb39b38ea
Parameters
learningRate : 0.3
randomJump : 99
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.11
Time: 2016-04-14 20:48:18
Commit: 7a59ad17fdf061d8ddaced90f71631e987e54524
Parameters
learningRate : 0.3
randomJump : 50
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.12
Time: 2016-04-14 15:41:25
Commit: 7d91669c2cbb094bab697f49bf4342c4a1a60476
Parameters
learningRate : 0.3
randomJump : 50
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.13
Time: 2016-04-14 21:23:09
Commit: 8a13707ad3055fad270c65049f3900b16d6a3009
Parameters
learningRate : 0.3
randomJump : 99
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.14
Time: 2016-04-14 17:58:19
Commit: 8c1e25edfd60117b4b7cd7ea175eb78a5b89d414
Parameters
learningRate : 0.3
randomJump : 99
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.15
Time: 2016-04-14 15:10:30
Commit: 9cdc90324e6ac7072e64557230c1d6fd0e7483d6
Parameters
learningRate : 0.3
randomJump : 1
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.16
Time: 2016-04-14 21:05:01
Commit: a71d39875ba0c1ad5a6b95ddbb19b479119d3161
Parameters
learningRate : 0.1
randomJump : 99
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.17
Time: 2016-04-14 17:06:24
Commit: bffb1423ade37619d965b5e3c836010242a6ed2d
Parameters
learningRate : 0.1
randomJump : 50
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.18
Time: 2016-04-14 17:27:02
Commit: c5350ce3ad8fe6d21e6926665624583793452c25
Parameters
learningRate : 0.3
randomJump : 50
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.19
Time: 2016-04-14 18:32:37
Commit: d8fa57941f047b3cf30aa65287563e85e9e5ab2c
Parameters
learningRate : 0.3
randomJump : 1
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.20
Time: 2016-04-14 16:30:11
Commit: dbe54ab9506c7d64a302c18e292d5e299113fb43
Parameters
learningRate : 0.1
randomJump : 1
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.21
Time: 2016-04-14 17:40:14
Commit: e7b49c0a2e89b87aecacf1eb7353760aa994e81c
Parameters
learningRate : 0.1
randomJump : 99
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : true
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.22
Time: 2016-04-14 19:07:48
Commit: ead43d435f9a5354c2167a315724eac906e36adb
Parameters
learningRate : 0.3
randomJump : 50
penaltyCalculatedOnlyOnce : true
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.23
Time: 2016-04-14 19:59:24
Commit: ef53c75162f1a1894493a4c580316debd2281dcd
Parameters
learningRate : 0.1
randomJump : 1
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
Experiment 7.24
Time: 2016-04-14 20:15:54
Commit: f4277af13f988ec185ca5823f868e8e0e9682975
Parameters
learningRate : 0.3
randomJump : 1
penaltyCalculatedOnlyOnce : false
respectKnowsLinksWeights : false
believe : 0.15
numberOfCycles : 150
numberOfAgents : 100
bufferSize : 10
Benefit per unit of time
</div>
</body> </html>
Experiment 6: Penalty Calculated Only Once Per Situation
Date: Thu Mar 17 15:37:54 2016
Analysis version (challprop-analytics): {analysis_commitNo}
Experiment version (challprop-experiments): 940d9ea7
Codebase version (challprop-java): c068ace4
(use git checkout $commit$ if needed)
Basic Statistics
These very roughly characterize what kind of data structure we get out of simulation.
##
## NUMBER OF VERTIXES:
## situation=11250
## benefitHolder=72566
## situationInst=83816
## agent=500
## god=1
##
## NUMBER OF EDGES:
## knowsLinks=6424
## challenged=207811
## interpreted=206966
## buffered=180683
## processed=72566
## reinterpreted=72566
## forgot=68411
##
## PERFORMANCE:
## Number of atomic events: 809003
## Total simulation time (secs): 21950.4182055690
## Total simulation time (hours): 6.0973383904
## Average time per atomic event (secs): 0.0271326784
Dynamics of benefit
##
## AVERAGE TOTAL BENEFIT:
## 14915.543698186015
##
## AVERAGE BENEFIT:
## [:]
Average benefit
## Warning: closing unused connection 6 (code.version)
Average benefit change per unit of time
Link distribution
Processes per agent
| eventsPerAgent | |
|---|---|
| originalChallenges | 22.50 |
| challenged | 415.62 |
| interpreted | 413.93 |
| buffered | 361.37 |
| processed | 145.13 |
| penalized | 22.50 |
| benefitted | 145.13 |
| forgot | 136.82 |
| reinterpreted | 145.13 |
Distribution of original challenges
## The number of original challenges in the simulation: 11250
Distribution of all challenges
Looks like normal distribution again. Note that is different from [Analysis 3/2](/blog/2014/12/26/analysis-3-slash-2-distribution-of-original-challenges/ where the average degree of an agent was much lower:
## Total number of all challenges in the simulation: 207811
Distribution of processed challenges
## Total number of all processed challenges in the simulation: 72566
Degree distribution
## Total number of outDegrees: 6424
Link weights
The number of links between agents in the graph: 6424; The average degree 2.14); The average link weight is 139.47.
## 1% 10% 20% 30% 50% 70% 90%
## 0.00001 0.00001 0.00001 0.00001 14.76612 141.36170 475.65542
## 99%
## 947.01738
0 of all links have negative or zero weight. A density plot of 98% of link weights (i.e. between 21 and 92) shows that they are more or less normally distributed:
Full simulation parameters
## $pathOnDisk
## [1] "/media/data/challprop-working/eln/"
##
## $generateTestDataOnDisk
## [1] "false"
##
## $tempConfigFile
## [1] "/media/challprop-working/temp/challprop.conf"
##
## $test
## [1] "false"
##
## $graphDatabase
## [1] "TitanCassandraLocal"
##
## $gephiVisualization
## [1] "false"
##
## $useDatabaseOfVectors
## [1] "true"
##
## $useDatabaseOfMatrixes
## [1] "true"
##
## $penaltyCalculatedOnlyOnce
## [1] "true"
##
## $rivalComponents
## [1] "[0, 1, 2, 3, 4]"
##
## $cstSituation
## [1] "0.05"
##
## $densityVectorSituation
## [1] "0.25"
##
## $percnegSituation
## [1] "0.2"
##
## $exponentSituation
## [1] "2"
##
## $numberOfAgents
## [1] "500"
##
## $decayFactor
## [1] "1"
##
## $decayRate
## [1] "0.03"
##
## $reciprocityRate
## [1] "0.2"
##
## $linkToGodWeight
## [1] "1.0"
##
## $weightImportance
## [1] "0.2"
##
## $believe
## [1] "0.15"
##
## $propagateRate
## [1] "0.4"
##
## $dimensions
## [1] "10"
##
## $densityMatrix
## [1] "0.5"
##
## $rfactor
## [1] "0.5"
##
## $numberOfCycles
## [1] "150"
##
## $learningRate
## [1] "0.1"
##
## $randomJump
## [1] "1"
##
## $propagationThreshold
## [1] "0"
##
## $disableRandomJump
## [1] "false"
##
## $maxBranchingFactor
## [1] "1000"
##
## $respectKnowsLinksWeights
## [1] "false"
##
## $bufferSize
## [1] "10"
##
## $ignoreNonRivalCorrection
## [1] "false"
##
## $cstNeed
## [1] "0.05"
##
## $densityVectorNeed
## [1] "0.5"
##
## $percnegNeed
## [1] "0.5"
##
## $exponentNeed
## [1] "2"
##
## $simulationStartWall
## [1] "32572676938179"
##
## $simulationStartCPU
## [1] "1291377944"
##
## $simulationFinishWall
## [1] "54523095143748"
##
## $simulationFinishCPU
## [1] "21095450486165"
##
## $archivingFinishedWall
## [1] "54642774669714"
##
## $archivingFinishedCPU
## [1] "21148847051074"
##
## $deletingCassandraWall
## [1] "54642795209516"
##
## $deletingCassandraCPU
## [1] "21148864190015"
Experiment 5: Less Agents in the Network
Date: Sat Mar 12 14:06:09 2016
Analysis version (challprop-analytics): {analysis_commitNo}
Experiment version (challprop-experiments): 2911856a
Codebase version (challprop-java): 77c6efd7
(use git checkout $commit$ if needed)
Basic Statistics
These very roughly characterize what kind of data structure we get out of simulation.
##
## NUMBER OF VERTIXES:
## situation=11250
## benefitHolder=73063
## situationInst=84313
## agent=500
## god=1
##
## NUMBER OF EDGES:
## knowsLinks=8622
## challenged=360144
## interpreted=358608
## buffered=229504
## processed=73063
## reinterpreted=73063
## forgot=116725
##
## PERFORMANCE:
## Number of atomic events: 1211107
## Total simulation time (secs): 7797.3883380440
## Total simulation time (hours): 2.1659412050
## Average time per atomic event (secs): 0.0064382324
Dynamics of benefit
##
## AVERAGE TOTAL BENEFIT:
## 17533.421462008027
##
## AVERAGE BENEFIT:
## [:]
Average benefit
## Warning: closing unused connection 6 (code.version)
Average benefit change per unit of time
Link distribution
Processes per agent
| eventsPerAgent | |
|---|---|
| originalChallenges | 22.50 |
| challenged | 720.29 |
| interpreted | 717.22 |
| buffered | 459.01 |
| processed | 146.13 |
| penalized | 717.22 |
| benefitted | 146.13 |
| forgot | 233.45 |
| reinterpreted | 146.13 |
Distribution of original challenges
## The number of original challenges in the simulation: 11250
Distribution of all challenges
Looks like normal distribution again. Note that is different from [Analysis 3/2](/blog/2014/12/26/analysis-3-slash-2-distribution-of-original-challenges/ where the average degree of an agent was much lower:
## Total number of all challenges in the simulation: 360144
Distribution of processed challenges
## Total number of all processed challenges in the simulation: 73063
Degree distribution
## Total number of outDegrees: 8622
Link weights
The number of links between agents in the graph: 8622; The average degree 2.87); The average link weight is 122.12.
## 1% 10% 20% 30% 50% 70%
## -70.017673 -37.782040 -14.650061 -1.533191 39.459057 128.027307
## 90% 99%
## 411.601103 942.087081
0.3436558 of all links have negative or zero weight. A density plot of 98% of link weights (i.e. between 21 and 92) shows that they are more or less normally distributed:
Full simulation parameters
## $pathOnDisk
## [1] "/media/data/challprop-working/eln/"
##
## $generateTestDataOnDisk
## [1] "false"
##
## $tempConfigFile
## [1] "/media/challprop-working/temp/challprop.conf"
##
## $test
## [1] "false"
##
## $graphDatabase
## [1] "TitanCassandraLocal"
##
## $gephiVisualization
## [1] "false"
##
## $useDatabaseOfVectors
## [1] "true"
##
## $useDatabaseOfMatrixes
## [1] "true"
##
## $rivalComponents
## [1] "[0, 1, 2, 3, 4]"
##
## $cstSituation
## [1] "0.05"
##
## $densityVectorSituation
## [1] "0.25"
##
## $percnegSituation
## [1] "0.2"
##
## $exponentSituation
## [1] "2"
##
## $numberOfAgents
## [1] "500"
##
## $decayFactor
## [1] "1"
##
## $decayRate
## [1] "0.03"
##
## $reciprocityRate
## [1] "0.2"
##
## $linkToGodWeight
## [1] "1.0"
##
## $weightImportance
## [1] "0.2"
##
## $believe
## [1] "0.15"
##
## $propagateRate
## [1] "0.4"
##
## $dimensions
## [1] "10"
##
## $densityMatrix
## [1] "0.5"
##
## $rfactor
## [1] "0.5"
##
## $numberOfCycles
## [1] "150"
##
## $newSituationsPerCycle
## [1] "5"
##
## $learningRate
## [1] "0.1"
##
## $randomJump
## [1] "1"
##
## $propagationThreshold
## [1] "0"
##
## $disableRandomJump
## [1] "false"
##
## $maxBranchingFactor
## [1] "1000"
##
## $respectKnowsLinksWeights
## [1] "false"
##
## $bufferSize
## [1] "10"
##
## $ignoreNonRivalCorrection
## [1] "false"
##
## $cstNeed
## [1] "0.05"
##
## $densityVectorNeed
## [1] "0.5"
##
## $percnegNeed
## [1] "0.5"
##
## $exponentNeed
## [1] "2"
##
## $simulationStartWall
## [1] "876386016069"
##
## $simulationStartCPU
## [1] "1197415854"
##
## $simulationFinishWall
## [1] "8673774354113"
##
## $simulationFinishCPU
## [1] "6714455343949"
##
## $archivingFinishedWall
## [1] "8957153230269"
##
## $archivingFinishedCPU
## [1] "6829430601420"
##
## $deletingCassandraWall
## [1] "8957167112575"
##
## $deletingCassandraCPU
## [1] "6829443668604"
Analysis 4.4/4 Degree Distribution
Date: Wed Feb 11 18:27:39 2015
Analysis version (challprop-analytics): 25a25dd
(use git checkout $commit$ if needed)
## Total number of outDegrees: 990
Analysis 4.4/3: Link Weights
Date: Wed Feb 11 18:12:00 2015
Analysis version (challprop-analytics): ee7d453
(use git checkout $commit$ if needed)
First, the number of links between agents in the graph: 990. The graph is sparse (average degree 0.33) given that we have 3000 of agents. The average link weight is 8.12. 98% of all link weights are between -21 and 92:
## 1% 10% 20% 30% 50% 70%
## -21.1730335 -2.3335033 -0.3338948 0.1175692 1.8546510 4.9825556
## 90% 99%
## 17.9730976 92.2649634
0.269697 of all links have negative or zero weight. A density plot of 98% of link weights (i.e. between 21 and 92) shows that they are more or less normally distributed:
This shows that most of the links are positive but very small
Analysis 4.5/4 Degree Distribution
Date: Wed Feb 11 17:29:52 2015
Analysis version (challprop-analytics): df5e83f
(use git checkout $commit$ if needed)
The number of out degrees of each agent is the same as in degrees, because we create reciprocal link for every link that is being created between two agents. Therefore we account only for outDegrees. The chart below shows that most of the agents have 10-15 outDegrees, but there are a few agents that have up to 30 connections to other agents.
## Total number of outDegrees: 13364
Ananysis 4.5/3 Link Weight Distribution
Date: Sat Dec 27 16:25:10 2014
Analysis version (challprop-analytics): 4cb1bdd
(use git checkout $commit$ if needed)
The number of links between agents in the graph: 13364. The graph is sparse (average degree 13.36) given that we have 1000 of agents. The average link weight is -9143.91. 98% of all link weights are between -16000 and -1480:
## 1% 10% 20% 30% 50% 70%
## -16004.864 -12997.479 -11814.849 -10957.176 -9448.605 -7733.493
## 90% 99%
## -4526.232 -1480.070
0.999551 (almost all) of all links have negative or zero weight. A density plot of 98% of link weights (i.e. between -16000 and -1480) shows that they are more or less normally distributed:
## stat_bin: binwidth defaulted to range/30. Use 'binwidth = x' to adjust this.
This shows that most of the links are positive but very small
Analysis 4.4/2 Distribution of Challenges
Date: Wed Feb 11 16:40:39 2015
Analysis version (challprop-analytics): {analysis_commitNo}
(use git checkout $commit$ if needed)
Processes per agent
| eventsPerAgent | |
|---|---|
| originalChallenges | 19.97 |
| challenged | 33.48 |
| interpreted | 33.29 |
| buffered | 32.93 |
| processed | 29.99 |
| penalized | 33.29 |
| benefitted | 29.99 |
| forgot | 18.33 |
| reinterpreted | 29.99 |
Distribution of original challenges
## The number of original challenges in the simulation: 60100
Distribution of all challenges
## Total number of all challenges in the simulation: 100775
Distribution of processed challenges
## Total number of all processed challenges in the simulation: 90278
Analysis 4.5/2: Distribution of Challenges
Date: Thu Feb 5 11:45:15 2015
Analysis version (challprop-analytics): 56f6334
(use git checkout $commit$ if needed)
Processes per agent
| eventsPerAgent | |
|---|---|
| originalChallenges | 21.60 |
| challenged | 647.19 |
| interpreted | 644.21 |
| buffered | 408.85 |
| processed | 139.29 |
| penalized | 644.21 |
| benefitted | 139.29 |
| forgot | 211.73 |
| reinterpreted | 139.29 |
Distribution of original challenges
The simulation goes this way: there are 10 thousand agents, 100 iterations. Each iteration 20% of agents get a fresh challenge generated by G-O-D, which is 2000 per iteration. These challenges in the beginning are (should be) distributed randomly across all agents in the network. If challenges are not distributed evenly across the network, that may cause buffers of agents lock some challenges and that would explain why we see somewhat low number of secondary propagation of challenges during further simulation. Lets see:
## The number of original challenges in the simulation: 21600
Distribution of all challenges
Looks like normal distribution again. Note that is different from [Analysis 3/2](/blog/2014/12/26/analysis-3-slash-2-distribution-of-original-challenges/) where the average degree of an agent was much lower:
## Total number of all challenges in the simulation: 647189
Distribution of processed challenges
## Total number of all processed challenges in the simulation: 139289
The number of processed challenges is almost identical for all agents. It is not surprising, because during each generation one agent can process only one challenge. This does not depend on how many challenges an agent gets. What is different, is that the more challenges an agent gets, the more ‘selection for relevance’ it performs.
A little discussion
Each original challenge is further propagated on average 29 times. Compare it to Experiment 3 where it is propagated on average 2.7 times.